Long-context understanding on RULER 1000 tokens
94.5NIAH
Steered Mamba
Updated 12d ago
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
2026.02 | 94.5 | 70 | 96 | 86.83 | |
2026.02 | 94 | 59 | 58 | 70.33 | |
2026.02 | 93.75 | 100 | 99 | 97.58 | |
2026.02 | 92 | 67 | 76 | 78.33 | |
2026.02 | 91.55 | 100 | 100 | 97.18 | |
2026.02 | 91 | 72 | 95.5 | 86.17 | |
2026.02 | 90.43 | 70 | 95 | 85.14 | |
2026.02 | 90 | 73 | 82 | 81.67 | |
2026.02 | 87.71 | 71 | 94 | 84.24 | |
2026.02 | 84 | 67 | 66.7 | 72.57 | |
2026.02 | 70.43 | 40 | 89 | 66.48 | |
2026.02 | 62 | 45 | 78 | 61.67 | |
2026.02 | 46 | 36 | 77 | 53 | |
2026.02 | 44.14 | 13.5 | 41 | 32.88 |